Fluctuation limits of site-dependent branching systems in critical and large dimensions (Q643228)

The author studies a branching model in which particles are initially distributed on \(\mathbb{R}^d\) according to a Poisson random field. The particles evolve independent of each other. Each particle moves in \(d\)-dimensional space according to a stochastic process \((\xi_1(t),\dots,\xi_d(t))_{t \geq 0}\). Each \(\xi_j\) is a symmetric \(\alpha_j\)-stable Lévy process (\(0 < \alpha_j \leq 2\)) and the coordinates are assumed to be independent of each other. Additionally, the particles split at rate \(\gamma > 0\) and generate offspring according to a site-dependent branching law with generating function \(g(s,x) = s + \sigma(x)(1-s)^2\), \(0 \leq s \leq 1\), \(0 \leq \sigma(x) \leq 1/2\), where \(x \in \mathbb{R}^d\) is the location parameter. The author denotes the empirical measure of the process at time \(s\) by \(N(s)\), that is, \(N(s)(A)\) equals the number of particles in \(A\) at time \(s\), \(A \subseteq \mathbb{R}^d\). The measure-valued process \(X(t) = \int_0^t (N(s)-\operatorname{E}N(s)) \, \mathrm{d}s\), \(t \geq 0\), is called the occupation time fluctuation. The re-scaled occupation time process \(X_n\) is defined by \[ \langle X_n(t), \phi \rangle = \frac{1}{F_n} \int_0^{nt} \langle N(s)-\operatorname{E}N(s), \phi \rangle \, \mathrm{d}s \] for appropriate test functions \(\phi\) and with a suitable scaling \(F_n\). The paper deals with functional limit theorems for \(X_n\) in the ``critical-dimension'' case \(\overline{\alpha} := \sum_{j=1}^{d} \alpha_j^{-1} = 2\) and in the ``large-dimension'' case \(\overline{\alpha} > 2\). In the critical-dimension case, the first of the main results states that, if \(\lim_{x \to \infty} \sigma(x) = \rho > 0\) and \(F_n^2 = n \log n\), then, for any smooth and rapidly decreasing function \(\psi: \mathbb{R}^{d+1} \to \mathbb{R}\) and any \(t > 0\), \(\int_0^t \langle X_n, \psi(\cdot,s) \rangle \mathrm{d}s\) converges in distribution to \(\int_0^t \langle X, \psi(\cdot,s) \rangle \mathrm{d}s\), where \(X\) is a centered Gaussian process. The covariance function is explicitly calculated in terms of the parameters of the model. The second main result covers the large-dimension case \(\overline{\alpha}\) and the critical-dimension case under the additional hypothesis that \(\int \sigma(x) \mathrm{d}x < \infty\). In this case, with \(F_n^2 = n\), \(\int_0^t \langle X_n, \psi(\cdot,s) \rangle \mathrm{d}s\) converges in distribution to \(\int_0^t \langle X, \psi(\cdot,s) \rangle \mathrm{d}s\) where \(X\) is again a centered Gaussian process but with a different covariance function. In particular, in the critical case, the scaling depends on the behavior of the function \(\sigma(x)\) that models the site-dependence of the branching law.