Brownian representation of a class of Lévy processes and its application to occupation times of diffusion processes (Q2505479)

The paper is devoted to Brownian representations of a class Lévy processes. First, the class \(M\) of Radon measures is introduced: \(M\) is the set of all functions \(m:(-\infty,\infty)\to (-\infty,\infty]\), such that \(m(x)=0\) on \((-\infty,0]\), \(m\) is non-decreasing, right-continuous on \((0,\infty)\), and \(\exists \delta>0\), s.t. \( \int_{0<x<\delta}m(x)^2\, dx<\infty. \) Further, put \[ G(x):=\int_0^xm(u)\, du,\;-\infty\leq x<l(m),\quad \text{and}\quad \zeta_{\pm}=\inf\{t\mid B(t)=\pm l(m)\}, \] where \(l(m)=\sup\{x: \, m(x)<\infty\}\), \(B(t)\) is the Brownian motion, and define the stochastic processes with the state space \(\mathbb{R}\cup\{\infty\}\): \[ S^{\pm}(m,t)= \begin{cases} \mp\int_0^t m(\pm B(s))dB(s)+G(B(t)),& 0\leq t<\zeta_\pm, \\ \infty,& t\geq \zeta_\pm, \end{cases} \] \[ T^{\pm}(m,t)=S^{\pm}(m,l^{-1}(t,0)),\quad t\geq 0, \] where \(l^{-1}(t,0)=\inf\{s>0\mid l(s,0)>t\}\) is the local time of \(B(t)\) w.r.t. \(2dx\). The authors prove that in case \(m(0+)>-\infty\) the following representations holds true: \[ S^{\pm}(m,t)=\int_{x>0}l(t,\pm x)dm(x)+m(0+)l(t,0),\quad 0\leq t<\zeta_\pm, \] \[ T^{\pm}(m,t)=\int_{x>0}l(l^{-1}(t,0),\pm x)dm(x)+m(0+)l(t,0),\quad 0\leq t<l(\zeta_\pm,0). \] Then in Section 2 under some conditions on the convergence of measures \(m_\lambda\) as \(\lambda\to 0\) it is proved that \[ (S^+(m_\lambda^+,t),S^-(m_\lambda^-,t),B(t))\longrightarrow (\sqrt{2} \sigma_+B^+(l(t,0)),\sqrt{2}\sigma_-B^-(l(t,0)),B(t)), \] \[ (T^+(m_\lambda^+,t),T^-(m_\lambda^-,t),B(t))\longrightarrow (\sqrt{2} \sigma_+B^+(l(t,0)),\sqrt{2}\sigma_-B^-(l(t,0)),B(t)), \] where \(\sigma_\pm\geq 0\), and the convergence is in law on the Skorokhod space of càdlàg functions, \(B^+\) and \(B^-\) are the copies of \(B(t)\), such that \(B^+\) and \(B^-\) are independent. In Section 3 the case of stable Lévy processes is treated. In particular, the convergence of renormalized \((S^\pm(m,t),B(t))\) and \((T^\pm(m,t),B(t))\) to, resp., \((S^\pm(m^{(\alpha)},t),B(t))\) and \((T^\pm(m^{(\alpha)},t),B(t))\) for suitable conditions on \(m\) is proved; here \(m^{(\alpha)}\), \(0<\alpha<2\), is 0 for \(x\leq 0\), and for \(x>0\) \[ m^{(\alpha)}=\begin{cases} x^{1/\alpha -1},& 0<\alpha<1, \\ \log x,& \alpha=1, \\ -x^{1/\alpha-1},& 1<\alpha<2. \end{cases} \] Also, as in Section 2, the central limit theorem for renormalized \((S^\pm(m,t),B(t))\) and \((T^\pm(m,t)\), \(B(t))\) is stated. In Section 4 the results from the preceding sections are applied to get limit theorems for occupation times of one-dimensional diffusion processes. Section 5 is devoted to a remark on Lamperti's class of distributions in connection with the results from Section 4. Section 6 is devoted to the properties of the Lévy measure of processes \(T^{\pm}(m,t)\).