[Paper Review] Central Limit Theorems for Super-OU Processes

Yan-Xia Ren, Renming Song+1 more|arXiv (Cornell University)|Feb 6, 2013

Stochastic processes and statistical mechanics9 references3 citations

TL;DR

This paper establishes central limit theorems for super-Ornstein-Uhlenbeck (super-OU) processes with general branching mechanisms under a second moment condition. Using backbone decomposition, it proves that normalized additive functionals converge to non-degenerate normal limits, sharpening prior results by ensuring non-degenerate asymptotic variance in all cases—critical, small, and large branching rate regimes—thereby resolving a key limitation in earlier work where limiting variances could vanish.

ABSTRACT

In this paper we study supercritical super-OU processes with general branching mechanisms satisfying a second moment condition. We establish central limit theorems for the super-OU processes. In the small and crtical branching rate cases, our central limit theorems sharpen the corresponding results in the recent preprint of Milos in that the limit normal random variables in our central limit theorems are non-degenerate. Our central limit theorems in the large branching rate case are completely new. The main tool of the paper is the so called "backbone decomposition" of superprocesses.

Motivation & Objective

• To establish central limit theorems for super-OU processes with general branching mechanisms satisfying a second moment condition.
• To address the limitation in prior work where limiting normal random variables in central limit theorems could be degenerate (i.e., zero variance).
• To provide a unified treatment across three regimes: small, critical, and large branching rates, with non-degenerate limits in all cases.
• To extend existing results by ensuring the limiting normal variables are non-degenerate, particularly in the large branching rate case, which was previously unaddressed.

Proposed method

• The backbone decomposition of superprocesses is used as the central analytical tool to decompose the superprocess into a branching particle system and a fluctuation component.
• The authors analyze the Laplace functional of the superprocess and derive a Feynman-Kac type representation for the moment generating function of additive functionals.
• They derive asymptotic expansions for the second moment of additive functionals, which depend on the sign of $\alpha - 2\gamma(f)b$, where $\gamma(f)$ is a spectral parameter related to the eigenfunction of the generator.
• The proof relies on martingale convergence and the use of the positive martingale $W_t = e^{-\alpha t}\|X_t\|$ converging almost surely and in $L^2$ to a non-degenerate limit $W_\infty$.
• A key step involves showing convergence in probability of normalized fluctuation terms, using moment bounds and dominated convergence arguments.
• The final convergence to a bivariate limit involving $W_\infty$ and a normal random variable is established via characteristic function convergence and the dominated convergence theorem.

Experimental results

Research questions

• RQ1 Can central limit theorems for super-OU processes be established with non-degenerate limiting normal distributions across all branching rate regimes?
• RQ2 Does the backbone decomposition technique allow for sharper results than previous approaches, particularly in eliminating degeneracy in the limiting variance?
• RQ3 Can the central limit theorem be extended to the large branching rate case, where no prior results existed?
• RQ4 How does the asymptotic variance of the normalized additive functional depend on the spatial structure and the branching mechanism?
• RQ5 Is the limiting normal variable in the central limit theorem independent of the total mass martingale $W_\infty$?

Key findings

• The limiting normal variable in the central limit theorem is non-degenerate for all branching rate regimes—small, critical, and large—resolving a key deficiency in earlier results where degeneracy could occur.
• In the large branching rate case, the paper establishes the first central limit theorem for super-OU processes, which was previously unknown.
• The asymptotic variance of the normalized additive functional is proportional to $\rho_f^2 W_\infty$, where $\rho_f^2 = A \sum_{|p| = \alpha/(2b)} a_p^2$, ensuring non-degeneracy when the coefficient sum is positive.
• The convergence of $e^{-\alpha t}\|X_t\|$ to $W_\infty$ holds almost surely and in $L^2$, guaranteeing that $W_\infty$ is non-degenerate and has finite second moment.
• The joint convergence of $ (e^{-\alpha t}\|X_t\|, C_t(\langle f,X_t\rangle - A_t)) $ to $ (W_\infty, \mathcal{N}(0, \rho_f^2 W_\infty)) $ is established, confirming the non-degenerate normal limit.
• The proof technique via backbone decomposition and characteristic function convergence ensures robustness and allows for precise control of higher-order fluctuations.