Abstract:In this paper, we consider a continuous-time Markov process and prove a local limit theorem for the integral of a time-inhomogeneous function of the process. One application is in the study of the fast-oscillating perturbations of linear dynamical systems.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the local limit theorem of the time integral of non - homogeneous Markov processes in time. Specifically, the author considered a continuous - time Markov process and proved the local limit theorem of the time integral of non - homogeneous functions in time in this process. One application of this result is in the study of rapidly oscillating perturbations of linear dynamical systems.
### Background and Problem Description of the Paper
In the paper, the author first defined a time - homogeneous Markov process \(\{X_s\}_{s\geq0}\) with a transition function \(P(s, x,\Gamma)\). Then, the author introduced some basic assumptions and notations, such as the Banach space \(B\) and the corresponding semigroup \(\{P_s\}_{s\geq0}\).
In order to state the local limit theorem, the author made some additional assumptions on the Markov process \(\{X_s\}_{s\geq0}\), among which the most important condition is that there exists a unique invariant measure \(\nu\) and \(X_s\) converges to this measure. This can be reformulated in terms of the quasi - compactness of the operator \(P_s\).
### Main Problems and Objectives
The main objective of the paper is to prove that there exists a local limit theorem for the time integral \(S_T=\int_{0}^{T}b(s/T,X_s)ds\) of the time - non - homogeneous function \(b:[0,1]\times X\rightarrow\mathbb{R}^d\). Specifically, the author hopes to prove:
\[ \lim_{T\rightarrow\infty}\left|\det(\Sigma)(2\pi T)^{d/2}E_{\mu}[f(X_T)g(S_T - u)]-e^{-\frac{1}{2}Tu^{*}(\Sigma\Sigma^{*})^{-1}u}\langle\nu,f\rangle\langle L,g\rangle\right| = 0\]
where:
- \(\Sigma\) is a positive definite matrix,
- \(g\in C(\mathbb{R}^d,\mathbb{R})\) is a compactly supported function,
- \(C > 0\) is a constant,
- \(f\in B_C\) is a real - valued function,
- \(\mu\in B_p'\) is the initial distribution,
- \(u\in\mathbb{R}^d\) is a vector.
### Applications
The paper also discussed an application of the local limit theorem, namely the averaging principle in stochastically perturbed linear dynamical systems. Specifically, consider the following equation:
\[ dY_t^{\varepsilon}=[A(t)Y_t^{\varepsilon}+v(t,X_{t/\varepsilon})]dt,\quad Y_0^{\varepsilon}=y\in\mathbb{R}^d\]
where \(A(t)\) is a continuously differentiable \(d\times d\) matrix function, \(v:[0,\infty)\times X\rightarrow\mathbb{R}^d\) is a continuous function, and \(\{X_s\}_{s\geq0}\) satisfies the above assumptions. Through the local limit theorem, it can be proved that:
\[ \lim_{\varepsilon\rightarrow0}\left|\det(\Sigma_t)(2\pi t/\varepsilon)^{d/2}E_{\mu}[f(X_{t/\varepsilon})g\left(\frac{1}{\varepsilon}(Y_t^{\varepsilon}-y_t)-u\right)]-e^{-\frac{1}{2}Tu^{*}(\Sigma_t\Sigma_t^{*})^{-1}u}\langle\nu,f\rangle\langle L,g\rangle\right| = 0\]