On the negativity of the top Lyapunov exponent for stochastic differential equations driven by fractional Brownian motion

Basic Information

• Paper ID: 2510.11531
• Title: On the negativity of the top Lyapunov exponent for stochastic differential equations driven by fractional Brownian motion
• Authors: Alexandra Blessing Neamţu, Mazyar Ghani Varzaneh
• Classification: math.PR (Probability Theory)
• Publication Date: October 14, 2025
• Paper Link: https://arxiv.org/abs/2510.11531

Abstract

This paper provides sign information for the top Lyapunov exponent of stochastic differential equations driven by fractional Brownian motion. To this end, the authors analyze the random dynamical systems generated by such equations, establish the random dynamical system structure, and construct appropriate invariant measures. By increasing the noise intensity and combining suitable density estimates with the Birkhoff ergodic theorem, they derive the negativity of the top Lyapunov exponent.

Research Background and Motivation

Problem Background

1. Dynamics of stochastic systems driven by non-Markovian noise: Traditional stochastic differential equation theory primarily addresses Markovian systems driven by Brownian motion, whereas systems driven by fractional Brownian motion exhibit long-range correlations and non-Markovian characteristics, resulting in more complex dynamical behavior.
2. Sign determination of Lyapunov exponents: Lyapunov exponents are crucial indicators for measuring stability and chaos in dynamical systems. A negative top Lyapunov exponent indicates system stability, while positive values suggest chaotic behavior or synchronization phenomena.
3. Absence of theoretical tools: In the Markovian case, the Fokker-Planck equation can be utilized to analyze invariant measures, but these tools no longer apply to systems driven by fractional Brownian motion.

Research Motivation

• Theoretical demand: Establish Lyapunov exponent theory for non-Markovian random dynamical systems
• Applied value: Provide stability analysis tools for fractional noise systems in fluid dynamics, neural networks, and other fields
• Methodological innovation: Develop new frameworks connecting stochastic dynamical systems (SDS) and random dynamical systems (RDS) theory

Core Contributions

1. First provision of sign information for Lyapunov exponents in non-Markovian settings: This is the first work establishing sign criteria for Lyapunov exponents of SDEs driven by fractional Brownian motion.
2. Establishment of connection between SDS and RDS theory: Construction of a transformation framework from stochastic dynamical systems to random dynamical systems, enabling the application of mature RDS theory to non-Markovian systems.
3. Construction of invariant measures and density estimates: Through decomposition theorems, invariant measures are constructed with Gaussian upper and lower bound estimates for their densities.
4. Proof that increased noise intensity leads to negative Lyapunov exponents: Main Theorem 4.1 demonstrates that by increasing noise intensity σ, the top Lyapunov exponent can be made negative.

Methodology Details

Problem Formulation

Consider the following stochastic differential equation driven by fractional Brownian motion:

dYt = F(Yt)dt + σdB^H_t
Y0 = x ∈ R^d

where B^H_t is a d-dimensional fractional Brownian motion with Hurst parameter H ∈ (0,1), and σ is the noise intensity parameter.

The objective is to analyze the sign of the top Lyapunov exponent λ^σ_1 of this equation, particularly to prove that λ^σ_1 < 0 when σ is sufficiently large.

Theoretical Framework

1. Mandelbrot-van Ness Representation of Fractional Brownian Motion

Fractional Brownian motion is constructed using a two-sided Wiener process:

B^H_t = (1/α_H) ∫_{-∞}^0 (-r)^{H-1/2}(dW(t+r) - dW(r))

2. Construction of Random Noise Process

A random noise process (B^H, {P_t}{t≥0}, P, {ϑ_t}{t≥0}) is defined on the Banach space B^H, where:

• Shift map: ϑ_t ω(s) := ω(s-t) - ω(-t)
• Transition operator: P_t(ω^-, ω^+) connects past and future noise components

3. Transformation from SDS to RDS

Theorem 3.2: Construction of random dynamical system Φ:

Φ^t_{(ω^-,ω^+)}(x) := φ^t_{P_t(ω^-,ω^+)}(x)

where φ is the original SDS. By defining an appropriate shift θ_t, the cocycle property is verified.

4. Construction of Invariant Measures

Theorem 3.11: Through decomposition theorems and completion arguments, a family of random measures {μ̃_{ω^-}} is constructed satisfying invariance:

(Φ^t_{(ω^-,ω^+)})_* μ̃_{ω^-} = μ̃_{P_t(ω^-,ω^+)}

Technical Innovations

1. Two-Component Noise Space

Innovatively employs Ω = B^H × B^H as the probability space, where the first component encodes the "past" of the noise and the second component encodes the "future," simultaneously satisfying SDS requirements and RDS framework.

2. Rescaling Technique

Through rescaling of the SDE:

dZ^σ_t = ||σ||^{-1}F(||σ||Z^σ_t)dt + ||σ||^{-1}σdB^H_t

the diffusion coefficient is standardized, facilitating analysis of the effect of parameter σ on invariant measures.

3. Wiener-Liouville Bridge Representation of Density

Using Proposition 5.16, the density of the invariant measure is represented as:

p̃^σ_{t_0}(l;y) = M(t_0,H,σ)exp(-||σ||^2|σ^{-1}(y-l(t_0))|^2/ρ^2_H t_0^{2H})G^σ_{t_0}(l;y)

where G^σ_ contains the expectation of the Wiener-Liouville bridge.

Experimental Setup

Theoretical Verification Framework

This work is primarily theoretical, with results verified through rigorous mathematical proofs rather than numerical experiments.

Key Assumptions

1. Assumption 4.1: σ is invertible, and F satisfies polynomial growth and monotonicity conditions
2. Assumption 5.1: F is eventually strictly monotone, i.e., there exists R > 0 such that for |ξ_1|,|ξ_2| ≥ R:

   ⟨F(ξ_2)-F(ξ_1), ξ_2-ξ_1⟩ ≤ -C^F_4|ξ_2-ξ_1|^2
   

Technical Conditions

• Hurst parameter H ∈ (0,1); when H ∈ (1/2,1), the derivative of F is required to be globally bounded
• Noise matrix σ belongs to the set T_{θ,κ}, controlling the condition number and minimum singular value

Main Results

Theorem 4.1 (Main Result)

Under Assumptions 5.1 and 5.3, for each θ ≥ 1, there exists a constant κ_θ > 0 such that for all σ ∈ T_{θ,κ_θ}, the top Lyapunov exponent of the SDE satisfies λ^σ_1 < 0.

Key Lemma 5.10

Upper bound estimate for the top Lyapunov exponent:

λ^σ_1 ≤ -C^F_4 π^σ(R^d\B(0,R)) + Cπ^σ(B(0,R))

Proposition 5.29 (Gaussian Bounds for Density)

For each bounded set B(0,R), there exist constants C_1,C_2 > 0 such that:

sup_{σ∈T_{θ,κ}} p̃^σ_∞(y) ≤ C_1 exp(-C_2|y|^2), ∀y ∈ B(0,R)

Proof Strategy

1. Step 1: Through Lemma 5.10, transform the negativity problem of the Lyapunov exponent into proving lim_{||σ||→∞}π^σ(B(0,R)) = 0
2. Step 2: Utilize rescaling techniques and density estimates to establish decay of probability mass of invariant measures on bounded sets
3. Step 3: Apply the Birkhoff ergodic theorem to complete the proof

Technical Challenges and Solutions

Challenge 1: Non-Markovian Nature

Solution: Through Mandelbrot-van Ness representation and two-component noise space design, embed the non-Markovian problem into a Markovian framework.

Challenge 2: Construction of Invariant Measures

Solution: Utilize decomposition theorems and completion techniques to construct families of random measures satisfying invariance on θ-invariant sets.

Challenge 3: Parameter-Dependent Density Estimates

Solution: Through Wiener-Liouville bridge representation and fractional calculus techniques, establish uniform density estimates with respect to parameter σ.

Challenge 4: Singularity Treatment for H > 1/2

Solution: In Proposition 5.26, eliminate singularities near zero by subtracting singular terms.

Related Work

Research in the Markovian Case

• Flandoli, Gess, Scheutzow (2017): Synchronization studies in the Brownian motion case
• Analysis of invariant measures using Fokker-Planck equations

Dynamics of Fractional Brownian Motion

• Hairer (2005): Establishment of SDS theory for fractional Brownian motion-driven SDEs
• Li, Panloup, Sieber (2023): Stationary density research for fractional-driven SDEs

Random Dynamical Systems Theory

• Arnold (1998): Foundational theoretical framework for RDS
• General theory of multiplicative ergodic theorems and Lyapunov exponents

Conclusions and Discussion

Main Conclusions

1. Theoretical breakthrough: First establishment of sign criteria for Lyapunov exponents in non-Markovian settings
2. Methodological innovation: Development of connection framework between SDS and RDS theory
3. Applied value: Provision of theoretical foundation for stability analysis of fractional noise systems

Limitations

1. Assumption conditions: Requirement that drift term F be eventually strictly monotone limits application scope
2. Technical complexity: Special treatment of singularities when H > 1/2 increases technical difficulty
3. Constructive nature: Main results are existential, without explicit expressions for κ_θ

Future Directions

1. Extension to gradient systems: Consider drift terms of the form F = -∇V
2. Multiplicative noise case: Study more general noise structures
3. Synchronization phenomena: Utilize negative Lyapunov exponents to study synchronization in fractional noise-driven systems
4. Numerical methods: Develop corresponding numerical algorithms to verify theoretical results

In-Depth Evaluation

Strengths

1. Theoretical innovation: Significant breakthrough in non-Markovian random dynamical systems
2. Methodological rigor: Precise mathematical reasoning and meticulous technical treatment
3. Framework generality: SDS-RDS connection framework has broad application potential
4. Problem importance: Addresses fundamental theoretical questions in stability analysis of fractional noise systems

Weaknesses

1. Application limitations: Currently applicable only to additive fractional noise and specific drift conditions
2. Computational complexity: Theoretical constructions are abstract, making practical computation difficult
3. Absence of numerical verification: Lacks numerical experiments validating theoretical predictions
4. Parameter dependence: Strong requirements on condition number of noise matrix

Impact

1. Academic contribution: Significant contribution to non-Markovian random dynamical systems theory
2. Application prospects: Potential applications in financial mathematics, biomathematics, and engineering control
3. Methodological value: SDS-RDS connection techniques generalizable to other non-Markovian systems

Applicable Scenarios

1. Theoretical research: Long-time behavior analysis of non-Markovian random dynamical systems
2. Engineering applications: Stability analysis of control systems with long-range correlated noise
3. Financial modeling: Risk assessment of financial models driven by fractional Brownian motion

References

This paper cites 58 related references, primarily including:

• Arnold (1998): Classical textbook on random dynamical systems
• Hairer (2005): Pioneering work on SDEs driven by fractional Brownian motion
• Li, Panloup, Sieber (2023): Recent density research for fractional-driven SDEs
• Flandoli, Gess, Scheutzow (2017): Important work on noise-induced synchronization

This report is based on in-depth reading of the full PDF text and accurately reflects the technical content and academic contributions of the paper. This work represents significant progress in non-Markovian random dynamical systems theory, laying a theoretical foundation for subsequent research in related fields.