Moderate deviations and laws of the iterated logarithm for the local times of additive Lévy processes and additive random walks (Q2370094)

Consider a \(d\)-dimensional symmetric Lévy process \(X(t) \) with characteristic exponent \(\psi ( \lambda ) \), that is, \(Ee^{t\lambda \cdot X( t) }=e^{-t\psi ( t) },\;t\geq 0,\;\lambda \in \mathbb{R}^{d}\). Assume there is a positive function \(a:\mathbb{R}_{+}\to \mathbb{R}_{+}\) such that \(\lim_{t\to \infty }a( t) =\infty \) and the limit \(\lim_{t\to \infty}t\psi ( \lambda/a( t)) =\Psi (\lambda )\) exists for any \(\lambda \in \mathbb{R}^{d}\). Then \(\Psi \) is continuous, nonnegative and \(\Psi ( r\lambda ) =r^{\alpha}\Psi ( \lambda ) ,\;r>0,\;\lambda \in \mathbb{R}^{d}\), where \(\alpha \in (0,2] \). Let \(p\geq 1\) be a fixed integer such that \(d<\alpha p\) and let \(X_{i}( t) ,1\leq i\leq p\), be \(p\) independent copies of \(X(t) \). The first goal of the author is to study the upper tail behaviour of the local time of the additive Lévy process \[ \overline{X}( t_{1},\dots ,t_{p}) =X_{1}( t_{1}) +\cdots +X_{p}( t_{p}) ,\quad t_{1},\dots ,t_{p}\in \mathbb{R}^{d}. \] The local time \(L( t,x) \) of \(\overline{X}( t_{1},\dots ,t_{p}) \) is defined differently in two different situations, namely, (i) \(\int_{\mathbb{R}^{d}}[ 1+\psi ( \lambda ) ] ^{-p}\,d\lambda <\infty ,\) and (ii) \(\psi ( \lambda +\mathbf k( 2\pi ) ) =\psi ( \lambda ),\;\mathbf k\in \mathbb Z^{d},\;\lambda \in \mathbb{R}^{d}\), as \[ \begin{aligned} L( t,x) &=\int_{[ 0,t] ^{p}}\delta _{x}( X_{1}( s_{1}) +\cdots +X_{p}( s_{p}) )\, ds_{1}\dots ds_{p},\quad t\geq 0,\;x\in \mathbb{R}^{d},\\ L( t,x) &=\int_{[ 0,t] ^{p}}\mathbf{1}_{\{ X_{1}( s_{1}) +\cdots +X_{p}( s_{p}) =x\} }\,ds_{1}\cdots ds_{p},\quad t\geq 0,\;x\in \mathbb{Z}^{d},\end{aligned} \] respectively. The second goal of the author is to study the upper tail behaviour of the local time of the additive random walk. Let \(S( n) \) be a symmetric \(d\)-dimensional random walk taking \(\mathbb{Z}^{d}\)-values with \( S( 0) =\mathbf{0}\). Assume that \(S( n) \) is in the domain of attraction of a nondegenerate \(d\)-dimensional stable law with characteristic exponent \(\Psi ( \lambda ) .\) Let the integer \( p\geq 1\) be fixed as before and let \(S_{1}( n) ,\dots ,S_{p}( n) \) be \(p\) independent copies of \(S( n) \). The multi-parameter process \(\overline{S}( n_{1},\dots ,n_{p}) =S_{1}(n_{1}) +\cdots +S_{p}( n_{p}) ,\;n_{1},\dots ,n_{p}=0,1,2,\dots ,\) is called the additive random walk generated by the random walks \(S_{1}( n) ,\dots ,S_{p}( n) \). Its local time \(\ell ( n,x) \) is defined as \[ \ell ( n,x)=\sum_{k_{1},\dots ,k_{p}=0}^{n}\mathbf{1}_{\{ S_{1}( k_{1})+\cdots +S_{p}( k_{p}) =x\} },\quad x\in \mathbb{Z} ^{d},\;n=1,2,\dots \;. \] The results established by the author are moderate deviations and laws of the iterated logarithm for \(L_{2}\)-norms of the local times and for the local times at a fixed site. The paper is a deep and very technical one. The results are stated in four theorems based on nine lemmas that are of interest in their own.