Scaling limits for symmetric Itô-Lévy processes in random medium (Q1045789): Difference between revisions

The aim of this paper is to obtain some functional limit theorems for the solutions of stochastic differential equations (SDE) driven by the Poisson random measures and Brownian motions. Let \((\Omega, G, \mu)\) be a probability space and \((\tau_x, x\in \mathbb{R})\) be a group of measure preserving transformations acting ergodically on \(\Omega\). So for any measurable function \(g\) on \((\Omega, G, \mu\) - the function \((x, \omega)\rightarrow g(\tau_X\omega)\) is measurable on \((R\times \Omega, B(R)\Theta G)\). The expectation with respect to the random medium is denoted by \(\mathbb{M}\). Let \(X\) be a solution of \((SDE)\) in the random environment \[ \begin{aligned} X_t=x + &\int^t_0(b+e)(\tau_{X_{r^-}}\omega)dr+\int^t_0\int_\mathbb{R}\gamma(\tau_{X_{r^-}}\omega, z)\widehat{N}(dr, dz)+\\ &\int^t_0\sigma(\tau_{X_{r^-}}\omega)dB_r,\end{aligned} \] where \(\widehat{N}\) is the compensated Poisson measure and \(B\) is Brownian motion. The coefficients in (1) are such that the generator of the process \(X\) for sufficiently smooth functions \(f\) (in a fixed environment \(\omega\)), coincide with \[ \begin{aligned} \mathcal{L}^\omega f(x) = &\frac{1}{2}e^{{2V(\tau_X\omega)}}\big(e^{-2V(\tau_X\omega)}a(\tau_x\omega)f'(x)\big )'\\ & +\lim\limits_{\epsilon\to 0}\int_{|z|>\epsilon}(f(x+z)-f(x))c(\tau_X\omega, z)e^{2V(\tau_X\omega)}\chi (dz),\end{aligned} \] where \(a, c, V\) are bounded functions of the environment and \(\chi (dz)\) is the Lévy measure related to the compensator of the Poisson random measure. The process \(X\) is diffusive enough and the matrix \(a=\sigma^2\) is uniformly elliptic. That implies the ergodicity of the model and allows to consider a large class of jump kernels \(c(\tau_x\omega, z)\). The Brownian motion, the Lévy process and the random medium are supposed independent. For each fixed \(\omega\), the symmetry and some other regularity conditions on the coefficients ensure the existence and pathwise uniqueness of the process \(X\). The main statement of the paper is Theorem 2.2. The two situations are possible. The pure jump scaling is in the case \(\int_{\mathbb{R}}z^2\chi (z)dz=+\infty\) in \(\mu\) probability, and then the rescaled process \(\mu (\varepsilon ) \cdot X (\frac{x}{\varepsilon})\) starting from \(0\in\mathbb{R}\) converges in law towards the Lévy process with \[ \int_\mathbb{R}(e^{iuz}-1-iuz{\mathbf{I}}_{|z|\leq 1}\mathbb{M}[\theta (\cdot \text{sign} (z)) \mathcal{H}(dz) \] in the Skorohod topology. It is required that \(\varepsilon^{-1}\delta^2(\varepsilon)\int\limits_{\delta(\varepsilon)|z|\leq\alpha}z^2\chi (dz)\rightarrow 0\) as \(\alpha\downarrow 0\) uniformly with respect to \(\varepsilon\); \(\lim\limits_{\varepsilon\rightarrow 0}\delta(\varepsilon ) =0\). Certain non-zero random function \(\theta :\{-1, 1\}\rightarrow L^{\infty}(\Omega )\) and a Lévy measure \(\mathcal{H}\) on \(\mathbb{R}\) are such that \[ \lim\limits_{\epsilon\rightarrow 0}\mathbb{M}\bigg [|\epsilon^{-1}\int_\mathbb{R}g(\delta (\epsilon )z)c(\cdot , z)\chi (dz) - \int_{\mathbb{R}}\theta (\cdot , \text{sign} (z))g(z)\mathcal{H}(dz)|\bigg ]=0 \] for each function \(g=\mathbf{I}_{[a, b]}\) with \(a<b\), \(0\notin [a, b]\). The diffusive scaling: in the case \(\int_\mathbb{R}z^2\chi (z) dz <+\infty\) in \(\mu\) probability, the rescaled process \(\epsilon^{\frac{1}{2}}X\big(\frac{x}{\epsilon}\big )\) starting from \(0\in \mathbb{R}\), converges in law in the Skorohod topology towards the non-standard centered Brownian motion with certain variance. In Section 2 the notations and the main statement Theorem 2.2 of the paper are stated. Here the definition of a random medium is reminded, the structure of the coefficients of equation (1) is discussed, several situations, where the main theorem can be applied, are considered. Examples with \(\alpha\)-stable, multi-stable kernels and with the kernels attracted by stable kernels are presented. The situation when the kernel \(c(\omega , z)\chi (dr)\) corresponds to that of a random walk is investigated. In Section~3 the Dirichlet forms in random medium are discussed. Section~4 deals with environment as seen from the particle and with the properties of the \(\Omega\)-valued process \(Y_t=\tau_{x_t}\omega\). In Section~5, Ergodic problems, the asymptotic properties of the process \(Y\) are considered. Section~6 deals with the problem of construction of correctors. Section~7 contains the tension estimates which are necessary to prove the functional theorems in Skorohod space. The main Theorem 2.2 is proved in Section~8. Several auxiliary lemmas used in different places of the paper are proved in the Appendix. The Introduction contains some discussion on the problem considered in the paper and on the earlier investigations in the field. The volume of the paper is 29 pages. The list of references contains 26 positions.