Model Reduction of Multivariate Geometric Brownian Motions and Localization in a Two-State Quantum System

Abstract: We develop a systematic framework for the model reduction of multivariate geometric Brownian motions, a fundamental class of stochastic processes with broad applications in mathematical finance, population biology, and statistical physics. Our approach highlights the interplay between the method of invariant manifolds and the adiabatic elimination procedure in deriving reduced equations in closed form for the deterministic part of the dynamics. An extended formulation of the fluctuation-dissipation theorem is also used to characterize the stochastic component of the reduced description. As a concrete application, we apply our reduction scheme to a geometric Brownian motion arising from a two-state quantum system, showing that the reduced dynamics accurately captures the localization properties of the original model while significantly simplifying the analysis.