[Paper Review] Large time behavior in random multiplicative processes

György Steinbrecher, X. Garbet+1 more|arXiv (Cornell University)|Jul 6, 2010

Mathematical Dynamics and Fractals11 references3 citations

TL;DR

This paper establishes a simple analytical formula for the critical exponent $\beta_c$ governing the large-time behavior of solutions to one-dimensional random multiplicative processes. Using novel weak convergence methods in generalized $L^p$ spaces and topological vector spaces, it proves that fractional moments $\mathbb{E}[|X_t|^p]$ remain bounded for $p < \beta_c$ and diverge for $p > \beta_c$, with convergence to a stationary state under a new, stronger topology than classical weak convergence.

ABSTRACT

In a general class of one dimensional random differential equation the convergence of the distribution function of the solution to stationary state distribution is studied. In particular it is proved the boundedness respectively the divergence of the fractional order moments of the solution below respectively above some critical exponent. This exponent is computed. In particular models it is the heavy tail exponent. When the equation is linear this exponent determines a new family of weak topologies (stronger compared to the classical one), related to the convergence to the stationary state.

Motivation & Objective

• To characterize the large-time behavior of solutions to a general class of one-dimensional random differential equations (RDEs) with stationary additive and multiplicative noise.
• To identify the critical exponent $\beta_c$ that separates bounded from divergent fractional order moments of the solution.
• To establish a new family of weak topologies—stronger than classical weak topology—under which the solution converges to a stationary distribution.
• To extend the analysis to nonlinear RDEs with weak nonlinearity, proving analogous moment behavior governed by the same $\beta_c$.
• To provide an explicit, analytically tractable formula for $\beta_c$ in terms of parameters of the multiplicative noise only, independent of the additive term.

Proposed method

• Formulates the solution of the linear RDE as $X_t = x_0 A_t + \widetilde{B}_t$, where $A_t$ and $\widetilde{B}_t$ are stochastic processes derived from the multiplicative and additive noise.
• Applies generalized $L^p$ spaces for $p \in (0,1)$ and $p > 1$, using the norm $\|f\|_p = \left(\mathbb{E}[|f|^p]\right)^{\sigma_p/p}$ to analyze moment behavior.
• Uses topological vector space methods to define a new weak convergence topology on probability measures, stronger than classical weak topology.
• Derives bounds on $\|X_t\|_p$ via inequalities (36)–(38), relating $\|X_t\|_p$ to $\|A_t\|_p$ and $\|\widetilde{B}_t\|_p$, with exponential growth rates $\gamma_p$.
• Establishes that $\gamma_p < 0$ for $p < \beta_c$ and $\gamma_p > 0$ for $p > \beta_c$, leading to bounded or divergent moments respectively.
• Extends the analysis to nonlinear RDEs via implicit integral forms and similar moment bounds, showing exponential divergence for $p > \beta_c$ and large initial conditions.

Experimental results

Research questions

• RQ1 What determines the critical exponent $\beta_c$ that separates bounded from divergent fractional order moments in random multiplicative processes?
• RQ2 How does the convergence to the stationary state occur, and what topological structure governs this convergence?
• RQ3 Can the critical exponent $\beta_c$ be computed explicitly from the parameters of the multiplicative noise, independent of the additive noise?
• RQ4 How does the behavior of fractional moments $\mathbb{E}[|X_t + z|^p]$ change for $p < \beta_c$ versus $p > \beta_c$?
• RQ5 To what extent can the results for linear RDEs be extended to nonlinear models with weak nonlinearity?

Key findings

• The critical exponent $\beta_c$ is given by a simple explicit formula (Equation 6) depending only on parameters of the multiplicative noise, not on the additive term.
• For $p < \beta_c$, the fractional moments $\mathbb{E}[|X_t + z|^p]$ remain bounded as $t \to \infty$, regardless of initial conditions.
• For $p > \beta_c$, the moments $\mathbb{E}[|X_t + z|^p]$ diverge exponentially for sufficiently large initial conditions.
• The convergence of the solution’s distribution to the stationary state occurs in a new, stronger weak topology than the classical one, defined via $C_\gamma(\mathbb{R})$ functions with $\gamma < \beta_c$.
• The solution $X_\infty$ belongs to $L^p$ for all $0 < p < \beta_c$, confirming the heavy-tailed nature of the stationary distribution with tail exponent $\beta_c$.
• In nonlinear RDEs with weak nonlinearity, the same $\beta_c$ governs moment behavior: bounded for $p < \beta_c$, divergent for $p > \beta_c$ when initial conditions are large.