Finite random iterated function systems do not always satisfy Bowen's formula

Basic Information

• Paper ID: 2509.02070
• Title: Finite random iterated function systems do not always satisfy Bowen's formula
• Author: Yuya Arima (Graduate School of Mathematics, Nagoya University)
• Classification: math.DS (Dynamical Systems)
• Publication Date: November 6, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2509.02070

Abstract

This paper constructs a finite random iterated function system satisfying the open set condition for which the random version of Bowen's formula fails. This counterexample demonstrates that analogous results established for random recursive constructions do not always apply to random iterated function systems.

Research Background and Motivation

Problem Background

1. Significance of studying random fractal sets: Random fractal sets, as models closer to natural phenomena than fractals generated by deterministic iterated function systems, hold important positions in mathematics and physics.
2. Two main random construction methods:
   • Random Iterated Function Systems (RIFSs): Uniformly select IFS at each level
   • Random Recursive Constructions: Independently select IFS for each finite string, exhibiting stronger independence
3. Limitations of existing theory: While dimension-theoretic results for random recursive constructions are known, it remains unclear whether the same results always hold for RIFSs.

Research Motivation

This paper aims to explore fundamental differences between RIFSs and random recursive constructions, particularly in fractal dimension theory. The author discovers that although both methods yield similar results in many cases, significant differences emerge in certain special cases.

Core Contributions

1. Construction of a key counterexample: Provides a finite random iterated function system satisfying the open set condition for which the random version of Bowen's formula fails
2. Revelation of theoretical differences: Demonstrates that results for random recursive constructions cannot be directly generalized to random iterated function systems
3. Rigorous mathematical proof: Provides complete mathematical argumentation by combining continued fraction theory and ergodic theory
4. Clarification of theoretical boundaries: Explicitly delineates the scope and limitations of RIFSs theory

Detailed Methodology

Task Definition

Construct a finite random iterated function system $(p⃗,Ψ)$ such that for $P$-almost all $ω ∈ Ω$: $\dim_H(J(Ψ(ω))) < B(Ψ)$ where $B(Ψ)$ is the Bowen parameter.

Construction Framework

1. Basic Setup

• Take $X = [0,1]^d$, define affine similarity maps: $φ_i(x) = \frac{1}{2}x + \frac{1}{2}v_i, \quad v_i = \sum_{ℓ=1}^d i_ℓe_ℓ$
• Define index sets:
  • $I_1 := \{0\}^d$ (singleton set)
  • $I_{2^d} := \{0,1\}^d$ (full vertex set)

2. Framework Definition

Introduce the concept of "framework" $F = (\{U_n\}_{n∈ℕ}, \{V_n\}_{n∈ℕ})$ satisfying:

• $(F1)$: $1 ≤ U_1$
• $(F2)$: $nU_n ≤ V_n$ and $(U_n + V_n)^3 ≤ U_{n+1}$

3. IFS Construction

For each $i ∈ ℕ$, define: $I^{(i)} := I_1^{U_i} × I_{2^d}^{V_i}$

Each $ψ_τ^{(i)}$ is a composition of $U_i + V_i$ maps.

Probability Vector Selection

Select probability vector: $p_n = \frac{1}{Cn^2}, \quad C = \sum_{n=1}^∞ \frac{1}{n^2}$

This choice echoes classical results in continued fraction theory.

Technical Innovations

1. Application of Continued Fraction Theory

Utilize probabilistic properties of continued fraction expansions, particularly extensions of Khinchin's theorem, to establish connections between random sequences and growth rates of continued fraction coefficients.

2. Clever Application of Ergodic Theory

Analyze limiting behavior through Birkhoff's ergodic theorem: $\lim_{n→∞} \frac{1}{n} \log \sum_{τ∈Σ_ω^n} (c_τ^{(ω)})^t = \begin{cases} ∞ & \text{if } t < d \\ -∞ & \text{if } t ≥ d \end{cases}$

3. Refined Analysis of Dimension Estimation

Through dimension theory of non-autonomous conformal IFS, prove that the Hausdorff dimension equals zero, creating a striking contrast with the Bowen parameter $d$.

Main Results

Theorem 1.1 (Main Result)

There exists a finite random iterated function system $(p⃗,Ψ)$ such that for $P$-almost all $ω ∈ Ω$: $\dim_H(J(Ψ(ω))) < B(Ψ)$

Proposition 2.2 (Bowen Parameter Calculation)

For the constructed system, $B(Ψ) = d$.

Proposition 2.4 (Dimension Result)

For $P$-almost all $ω ∈ Ω$, $\dim_H(J(Ψ(ω))) = 0$.

Proof Strategy

Step One: Bowen Parameter Analysis

Prove through direct calculation: $E_{i∈ℕ}\left[\log \sum_{j∈I^{(i)}} (c_j^{(i)})^t\right] = \begin{cases} ∞ & \text{if } t < d \\ -∞ & \text{if } t ≥ d \end{cases}$

Step Two: Continued Fraction Properties

Utilize the correspondence between probability vector $p_n = 1/(Cn^2)$ and the distribution of continued fraction coefficients to construct random sequences with special growth properties.

Step Three: Upper Bound Estimation of Dimension

Apply the dimension formula for non-autonomous systems: $\dim_H(J(Φ_ω)) ≤ \inf\{t ≥ 0: P(t) < 0\}$ where $P(t) = \liminf_{n→∞} \frac{1}{n} \log \sum_{τ̃∈Σ̃_ω^n} c_{τ̃}^t$

Related Work

Random Recursive Construction Theory

Mauldin-Williams (1986) proved that for random recursive constructions, the Hausdorff dimension almost surely equals: $\inf\left\{t ≥ 0: \log \sum_{i∈ℕ} p_i \sum_{j∈I^{(i)}} (c_j^{(i)})^t ≤ 0\right\}$

Known Results for RIFSs

Roy-Urbański (2011) and Rempe-Gillen-Urbański (2016) proved Bowen's formula for RIFSs under specific conditions.

Breakthrough of This Paper

This paper is the first to demonstrate the necessity of these conditions, revealing the essential limitations of RIFSs theory.

Conclusions and Discussion

Main Conclusions

1. Clarification of theoretical boundaries: RIFSs do not always satisfy Bowen's formula, exhibiting essential differences from random recursive constructions
2. Effectiveness of construction method: Successfully separates two concepts of randomness through clever probability distribution design and framework construction
3. Refinement of dimension theory: Provides more precise understanding of dimension theory for random fractals

Theoretical Significance

The results indicate that the "strength" of randomness has fundamental effects on fractal dimension. Independent selection at each position in random recursive constructions versus level-wise uniform selection in RIFSs leads to essentially different geometric properties.

Limitations

1. Specificity of construction: The counterexample depends on particular probability distributions and geometric configurations
2. Extremality of zero dimension: The constructed limit set has dimension zero, which may not be sufficiently general
3. Higher-dimensional cases: Theory remains incomplete for general dimensional cases

Future Directions

1. Finding general conditions: Determine necessary and sufficient conditions for RIFSs to satisfy Bowen's formula
2. Intermediate dimension cases: Construct examples with dimension between zero and the Bowen parameter
3. Extension of applications: Generalize results to more general random dynamical systems

In-Depth Evaluation

Strengths

1. Significant theoretical contribution: First to reveal essential differences between two important random construction methods
2. Sophisticated technical methods: Cleverly combines probability theory, ergodic theory, and fractal geometry
3. Rigorous and complete proof: Mathematical argumentation is logically clear with proper handling of technical details
4. Profound problem formulation: Addresses core theoretical questions in the field

Weaknesses

1. Specificity of counterexample: Construction depends on particular parameter choices; generality needs strengthening
2. Insufficient geometric intuition: Geometric explanation for why such differences occur is inadequate
3. Unclear application prospects: Practical application value of theoretical results requires further exploration

Impact

1. Theoretical value: Provides important counterexample and new research directions for random fractal theory
2. Methodological contribution: Demonstrates potential applications of continued fraction theory in random fractals
3. Field advancement: Promotes RIFSs theory toward more precise and complete development

Applicable Scenarios

The methods and results of this paper apply to:

1. Theoretical research in random fractal geometry
2. Dimension analysis of random dynamical systems
3. Interdisciplinary research between probability theory and fractal geometry
4. Theoretical foundations of random fractal models in natural phenomena

References

The paper cites important literature in the field, including:

• Mauldin & Williams (1986): Foundational work on random recursive constructions
• Roy & Urbański (2011): Random graph-directed Markov systems
• Rempe-Gillen & Urbański (2016): Theory of non-autonomous conformal IFS
• Khinchin (1964): Classical textbook on continued fraction theory

This paper makes important contributions to random fractal theory, not only resolving an important theoretical question but also providing new research directions and technical tools for further development in the field.