Abstract:The present paper is devoted to the investigation of an important family of absorbed singular diffusion processes exhibiting long transient dynamics, namely, interesting and important dynamical behaviours over long but finite time scales. We explore the multiscale dynamics by establishing the asymptotic distribution of the normalized extinction time, the asymptotic reciprocal relationship between the mean extinction time and the principal eigenvalue of the generator, and a sophisticated multiscale estimate of solutions. While information about the extinction time and mean extinction time uncovers fundamental principles quantifying in particular the lifespan of interesting dynamical behaviours combined and its natural connection with the principal eigenvalue, the multiscale estimate characterizes the dynamics over different time scales. These are achieved by examining quasi-stationary distributions (QSDs) that govern the dynamics before the eventual absorption happens, and establishing the powerful sub-exponential large deviation principle (LDP) for QSDs, which determines the quasi-potential function and prefactor in the WKB expansion, and therefore, provides very fine asymptotic or concentration properties of QSDs. To the best of our knowledge, this is the first time that the sub-exponential LDP for QSDs is established for absorbed singular diffusion processes. Our approach is analytic and elementary. As byproducts or consequences, new results about QSDs near the absorbing state and infinity, the sub-exponential asymptotic of the principal eigenvalue, and the asymptotic of the principal eigenfunction are obtained. The sub-exponential LDP for QSDs is of independent interest and expected to have more far-reaching consequences. Applications to logistic diffusion processes arising from chemical reactions and population dynamics are discussed. In particular, Keizer's paradox concerning the long-term dynamical disagreement between a deterministic model and its stochastic counterpart, and diffusion approximation of QSDs are rigorously justified.

Large deviations and related problems for absorbing Markov chains

Large Deviations for Markov Chains and Spectrum of Matrices

Precise large deviations for randomly weighted sums of negatively dependent random variables with consistently varying tails

Large Deviations for Empirical Measures of Self-Interacting Markov Chains

Large Deviations for Inhomogeneous Systems

Large Deviations for Infinite Dimensional and Reversible Reaction-Diffusion Processes

ON THE LARGE DEVIATION FUNCTIONS OF MARKOV CHAINS

Large deviations for macroscopic observables of heavy-tailed matrices

Large Deviations for Stochastic Differential Delay Equations with Lévy Noises

Large Deviations for SDE driven by Heavy-tailed Lévy Processes

Large Deviations for Stochastic Differential Equations Driven by Semimartingales

Large Deviation Principles of Invariant Measures of Stochastic Reaction-Diffusion Lattice Systems

Large Deviations for Empirical Processes of Interacting Particle Systems

Large deviation principles for markov processes via Φ-sobolev inequalities

Large deviation principle for quasi-stationary distributions and multiscale dynamics of absorbed singular diffusions

Large deviations for metastable states of Markov processes with absorbing states with applications to population models in stable or randomly switching environment

Large deviations asymptotics for large waiting times

Large Deviations for Exchangeable Observations with Applications

Large deviations for invariant measures of multivalued stochastic differential equations with jumps

Large deviations for some dependent heavy tailed random sequences

Limiting Process of Absorbing Markov Chains