Conditional speed of branching Brownian motion, skeleton decomposition and application to random obstacles (Q2357272)

The paper deals with \(d\)-dimensional branching Brownian motion starting with one initial particle at the origin at time \(0\). The particle performs a Brownian motion for some time which is given by the value of an independent exponential random variable with parameter \(\beta>0\). At the time of its death, the particle gives birth to an independent random number of offspring distributed according to some probability measure \(\lambda\) on \(\mathbb{N}_0 = \{0,1,2,\ldots\}\). Analogously, each offspring particle repeats the same procedure independently of all others starting from the final position of her parent. Let \(Z(t)\) be the point process of the positions of all individuals in the system at time \(t\). Then \((Z(t))_{t \geq 0}\) is a measure-valued Markov process called branching Brownian motion. The authors study \((Z(t))_{t \geq 0}\) in an Poissonian environment of traps. More precisely, let \(\Pi\) denote a Poisson point process on (the Borel sets of) \(\mathbb{R}^d\) with intensity measure \(\nu\) which is absolutely continuous with respect to \(d\)-dimensional Lebesgue measure and possesses a continuous density \(f\) satisfying \[ f(x) \sim \frac{l}{|x|^{d-1}} \quad \text{as } |x| \to \infty \] where \(l>0\). For \(a>0\), let \(K = \bigcup_{x \in \Pi} \bar B(x,a)\) where summation is over the points of \(\Pi\) and \(\bar B(x,a)\) is the closed ball with radius \(a\) centered around \(x\). The aim of the paper is to study the annealed probability that Brownian motion avoids the trap \(K\) at least until time \(t\). More precisely, if \(\mathbb{P}\) denotes the law of the Poisson process \(\Pi\) and \(\mathbb{E}\) the corresponding expectation, and if \(\mathrm{P}\) is the law of branching Brownian motion, then they study the annealed probability \((\mathbb{E} \times \mathrm{P})(T>t)\) where \(T\) is the infimum of times \(t\) for which \(\mathrm{supp}(Z(t)) \cap K\) is nonempty. In their main result, the authors give an expression in terms of the parameters of the model for \(\lim_{t \to \infty} \frac1t \log (\mathbb{E} \times \mathrm{P})(T>t)\) in the case \(\lambda(0)=0\) (each particle has at least one offspring). The case \(\lambda(0)>0\), in which extinction is possible, is divided into two subcases. In the supercritical case, the authors provide a result analogous to the one for the case \(\lambda(0)=0\) but for the annealed probability \((\mathbb{E} \times \mathrm{P})(T>t | \mathcal{E}^c)\) where \(\mathcal{E}\) is the event that the branching process dies out. Finally, in the critical and subcritical case, in which extinction is certain, they show \(\lim_{t \to \infty} (\mathbb{E} \times \mathrm{P})(T>t) = (\mathbb{E} \times \mathrm{P})(T > \tau)\) where \(\tau\) is the time of extinction. To prove their main result, the authors prove a theorem on the speed of branching Brownian motion conditioned on non-extinction, which is of interest in its own right. The proof of the latter theorem relies on a skeleton decomposition of the branching Brownian motion.