A long range dependence stable process and an infinite variance branching system (Q2371946)

The authors' object of study is a \((d, \alpha, \beta)\)-branching particle system in continuous time, that is a system of particles moving independently in \(\mathbb{R}^d\) as symmetric \(\alpha\)-stable Lévy processes, whose branching law is given and lies in the domain of attraction of a \((1+\beta)\)-stable law, \(0<\beta<1\), if the initial state is a uniform Poisson random measure, i.e., the intensity measure equals the Lebesgue measure. While this branching system and its associate superprocess have been studied for a long time, the interest of this work lies in the asymptotic behaviour of the occupation time fluctuation. The first main result is a functional limit theorem for the appropriately rescaled occupation time fluctuations \(X_T\) of the system, \(\alpha <2\), \(\beta <1\) and dimensions \(\alpha/\beta < d < \alpha (1+\beta)/\beta\) towards a process \(K\lambda \xi\) for \(T\rightarrow \infty\), where \(K\) is a (random) constant, \(\lambda\) is the Lebesgue measure on \(\mathbb{R}^d\), and \((\xi_t)_{t\geq 0}\) is an explicitely known \((1+\beta)\) stable-process, which exhibits long range dependence. This is an analogous phenomenon to the case of \(\alpha <2\), \(\beta=1\) and \(\alpha < d < 2\alpha\), where \((\xi_t)_{t\geq 0}\) turns out to be a so called sub-fractional Brownian motion, which is a special Gaussian process, also equipped with this property. The mode of convergence takes place weakly in \(\mathcal{C}([0,t], \mathcal{S}'(\mathbb{R}^d))\) and is obtained by mainly analytical arguments exploiting topological properties of \(\mathcal{S}'(\mathbb{R}^d)\). The second main result consists in the further investiation of the limiting process in time \((\xi_t)_{t\geq 0}\) and its long range dependence property. The long range dependence is characterized by a dependence exponent \(\kappa\), which describes the asymptotic behaviour of the codifferences of increments of \(\xi\) on time intervals far apart from each other. In the present case the authors identify two long range dependence regimes, one for \(\beta > d/(d+\alpha)\), where \(\kappa = d/\alpha\) and hence coincides with the case of binary, finite variance branching and a second one for \(\beta \leq d/(d+\alpha)\), \(\kappa = d/\alpha (1+\beta-d/(\alpha + d))\), now explicitly depending on the value of \(\beta\).