Existence of Invariant Measures for Delay Equations with Stochastic Negative Feedback

Basic Information

• Paper ID: 2501.00141
• Title: Existence of Invariant Measures for Delay Equations with Stochastic Negative Feedback
• Authors: M. van den Bosch (Leiden University), O.W. van Gaans (Leiden University), S.M. Verduyn Lunel (University of Utrecht)
• Classification: math.DS (Dynamical Systems), math.PR (Probability Theory)
• Publication Date: January 3, 2025
• Paper Link: https://arxiv.org/abs/2501.00141

Abstract

This paper provides sufficient conditions for the existence of invariant probability measures for general stochastic differential equations with finite time delays. Applications include the Mackey-Glass equation and Nicholson's blowflies equation perturbed by (small) multiplicative noise terms. Solutions of these stochastic negative feedback systems exist globally and remain bounded in the probabilistic sense. The study reveals that the occurrence of finite-time explosions is closely related to the boundedness of solutions and solution segments in the probabilistic sense. Non-trivial invariant measures exist if and only if there exists at least one initial condition such that the solution remains bounded away from zero in the probabilistic sense. The noise driving the dynamical system can be an integrable Lévy process.

Research Background and Motivation

Problem Description

This paper investigates the existence of invariant measures for delay differential equations with stochastic negative feedback. The core objects of study are deterministic delay differential equations of the form:

x'(t) = -γ(t)x(t) + r(t)f(x(t-τ))

and their stochastic perturbations.

Research Significance

1. Biological Relevance: The Mackey-Glass equation models white blood cell concentration, while Nicholson's blowflies equation describes insect population dynamics. These models have important applications in physiology and ecology.
2. Mathematical Challenges: Chaotic behavior in delay differential equations has not yet been rigorously proven. Finding invariant measures provides a new avenue for understanding the long-term behavior of these systems.
3. Stochastic Perturbations: Real systems are inevitably subject to noise, making the study of stochastic versions more relevant to practical applications.

Limitations of Existing Methods

• For the Mackey-Glass equation and Nicholson's blowflies equation, chaotic phenomena have not yet been mathematically proven.
• Existing results on global attractors only apply to parameter regions that allow steady-state and periodic solutions.
• There is a lack of systematic approaches for handling invariant measures of general stochastic delay differential equations.

Core Contributions

1. Establishment of a General Theoretical Framework: Provides sufficient conditions for the existence of invariant measures for stochastic differential equations with finite time delays.
2. Proof of Global Existence of Solutions: All solutions exist globally and remain bounded in the probabilistic sense.
3. Characterization of Conditions for Invariant Measure Existence: Non-trivial invariant measures exist if and only if there exists at least one initial condition such that the solution remains bounded away from zero in the probabilistic sense.
4. Extension of Noise Types: Allows noise to be an integrable Lévy process, not limited to Brownian motion.
5. Application to Classical Models: Proves the existence of invariant measures for stochastic versions of the Mackey-Glass equation and Nicholson's blowflies equation.

Methodology Details

Problem Formulation

Study autonomous stochastic delay differential equations:

dX(t) = a(X_t) dt + b(X_{t-}) dM(t), t ≥ 0
X(u) = Φ(u), u ∈ [-τ, 0]

where M(t) is a semimartingale, a and b are locally Lipschitz functionals, and X_t denotes the solution segment process at time t.

Core Methodological Framework

1. Change of Variables Technique

Through logarithmic transformation Y(t) = log X(t), the original equation is transformed to:

dY(t) = [-γ(t) + r(t)e^{-Y(t)}f(e^{Y(t-τ)})] dt + a(Y_t,t) dt + b(Y_t,t) dW(t)

This transformation preserves solution positivity and facilitates analysis.

2. Path Estimation Methods

• Upper Bound Estimates: Utilizing the negative feedback property, when the solution is sufficiently large, the drift term becomes negative, preventing solutions from diverging to infinity.
• Lower Bound Estimates: When f(0) > 0, integral inequality techniques are employed to obtain lower bounds for solutions.

3. Probabilistic Boundedness Analysis

Establishes connections among three key concepts:

• Global existence
• Probabilistic boundedness of solutions
• Probabilistic boundedness of solution segments

4. Compactness Conditions

Through the Arzelà-Ascoli theorem and Skorokhod topology, sufficient conditions for solution segment compactness are established:

• Condition (i): Noise coefficients are bounded
• Condition (ii): The supremum of drift terms is bounded in the probabilistic sense
• Condition (iii): The supremum norm of solution segments is bounded in the probabilistic sense

Technical Innovations

1. Backward-Time Supremum Technique

For Lévy-driven processes with negative drift:

Y(t) = -∫₀ᵗ a(s) ds + ∫₀ᵗ b(s) dL(s)

It is proven that the backward-time supremum process (sup_{0≤θ≤t}(Y(t)-Y(θ)))_{t≥0} is bounded in the probabilistic sense.

2. Extension of the Krylov-Bogoliubov Method

The classical Krylov-Bogoliubov existence theorem is extended to the space of right-continuous left-limited functions D-τ,0, handling jump noise cases.

3. Application of Semimartingale Characteristics

Semimartingale local characteristics (B^X, C^X, ν^X) are utilized to prove the compactness of solution segment processes, applicable to a broader class of noise types.

Experimental Setup

Theoretical Verification

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

Application Models

1. Mackey-Glass Equation:

   x'(t) = -γx(t) + r·x(t-1)/(1+x(t-1)^p), p > 1
   

2. Nicholson's Blowflies Equation:

   x'(t) = -γx(t) + r·x(t-1)e^{-px(t-1)}, p > 0
   

Noise Types

• Brownian motion
• Lévy processes with bounded jumps
• Integrable Lévy processes

Main Results

Theorem 1.1 (Main Result)

Assume f: ℝ → ℝ is locally Lipschitz continuous, non-negative and bounded on (0,∞), inf_{t≥0} γ(t) > 0, sup_{t≥0} r(t) < ∞. Then the stochastic delay differential equation:

dX(t) = [-γ(t)X(t) + r(t)f(X(t-τ))] dt + σX(t)c(X_t) dW(t)

has a unique solution that exists globally and remains bounded in the probabilistic sense for almost all non-negative F₀-measurable initial values.

Furthermore, when γ and r are constants:

• If f(0) > 0, then almost all solutions remain bounded away from zero in the probabilistic sense, and there exists a stationary distribution μ supported on (0,∞).
• If f(0) = 0, then there exist two distinct stationary distributions: the Dirac measure δ₀ and a stationary distribution μ supported on (0,∞) (provided that there exists at least one initial condition such that the solution remains bounded away from zero).

Corollary 1.2 (Mackey-Glass Equation)

For the deterministic Mackey-Glass equation, when p > 1, inf_{t∈-τ,0} φ(t) > 0, lim inf_{t→∞} r(t)/γ(t) > 1, solutions remain bounded away from zero. When γ and r are constants and r > γ, the equation admits a non-trivial invariant measure on C-1,0.

Related Work

Delay Differential Equation Theory

• Classical theory by Hale & Verduyn Lunel provides the foundational framework for this work.
• Related work on Wright equations provides precedent for stochastic perturbations.
• Research on global attractors provides reference for asymptotic behavior analysis.

Stochastic Differential Equations

• Stochastic differential equation theory by Mao
• Semimartingale integration theory by Protter
• Work on ergodicity of infinite-dimensional systems by Da Prato & Zabczyk

Invariant Measure Theory

• Classical Krylov-Bogoliubov existence theorem
• Research on Feller properties of finite-delay systems
• Compactness and ergodicity theory

Conclusions and Discussion

Main Conclusions

1. A general theory for the existence of invariant measures for stochastic negative feedback delay equations has been established.
2. Global existence and probabilistic boundedness of solutions have been proven.
3. Necessary and sufficient conditions for the existence of non-trivial invariant measures have been characterized.
4. A rigorous mathematical foundation has been provided for classical biomathematical models.

Limitations

1. Case f(0) = 0: When f(0) = 0, proving that solutions remain bounded away from zero in the probabilistic sense remains an open problem.
2. Uniqueness: Uniqueness of invariant measures has not been proven.
3. Computational Aspects: Effective numerical methods for computing invariant measures are lacking.
4. High-Dimensional Extensions: Extension of the method to system cases requires further work.

Future Directions

1. Study lower bound estimates for solutions when f(0) = 0.
2. Explore uniqueness, regularity, and absolute continuity of invariant measures.
3. Develop numerical methods for computing invariant measures.
4. Extend to multidimensional delay systems.
5. Study asymptotic stability and convergence rates.

In-Depth Evaluation

Strengths

1. Theoretical Innovation: Successfully extends the Krylov-Bogoliubov method to stochastic delay differential equations with high technical difficulty.
2. Systematic Methodology: Establishes a complete theoretical chain from local existence to global existence, from probabilistic boundedness to compactness, and finally to invariant measure existence.
3. Application Value: Provides a rigorous theoretical foundation for important biomathematical models.
4. Technical Depth: The application of backward-time supremum techniques and semimartingale characteristics demonstrates sophisticated technical mastery.

Weaknesses

1. Completeness: The treatment of the f(0) = 0 case is incomplete, representing a significant theoretical gap.
2. Practicality: Lacks numerical verification and computational methods, making theoretical results difficult to verify in practice.
3. Generalizability: The method is primarily applicable to scalar equations, with extension to systems requiring additional work.

Impact

1. Academic Contribution: Makes important contributions to stochastic delay differential equation theory, filling theoretical gaps in the field.
2. Application Prospects: Provides new tools for analyzing stochastic models in biology, ecology, and related fields.
3. Methodological Value: The developed techniques have important reference value for related research.

Applicable Scenarios

This paper's methods are applicable to:

• Modeling biological systems with negative feedback mechanisms
• Analyzing noise-perturbed delay dynamical systems
• Systems requiring investigation of long-term behavior
• Delay differential equations with multiplicative noise

References

The paper cites 88 important references covering classical and cutting-edge work in delay differential equations, stochastic analysis, ergodic theory, and related fields, demonstrating solid theoretical foundations and broad academic perspective.