Derivative martingale of the branching Brownian motion in dimension \(d\ge 1\) (Q2077343): Difference between revisions

Consider a binary branching Brownian motion $(X_t(j): j=1,\ldots,\mathcal{N}_t)_{t \geq 0}$ in dimension $d \in \mathbb{N}$, where $\mathcal{N}_t$ is the number of particles at time $t$ and $X_t(j)$ is the position in $\mathbb{R}^d$ of the $j$th particle alive at time $t$. Interest lies in the maximal displacement from the origin: \[ \max_{j=1,\ldots,\mathcal{N}_t} |X_t(j)|, \quad t \geq 0 \] where $|\cdot|$ denotes the Euclidean distance in $\mathbb{R}^d$. \textit{B. Mallein} [Electron. Commun. Probab. 20, Paper No. 76, 12 p. (2015; Zbl 1329.60307)] showed that \[ R_t = 2t + \frac{d-4}{2\sqrt{2}} \log t + Y_t \quad \text{as } t \to \infty \] where the remainder term $Y_t$ is a stochastic process such that $\sup_{t \geq 1} |Y_t|$ is an a.\,s.\ finite random variable. In one dimension, the fine behavior of the maximal displacement is governed by the derivative martingale. In dimension $d$, there is a family of analogues, namely, for every $\theta \in \mathbb{S}^{d-1}$, \[ Z_t(\theta) = \sum_{j \in \mathcal{N}_t} (\sqrt{2}t - X_t(j) \cdot \theta) e^{X_t(j) \cdot \theta - \sqrt{2}t}, \quad t \geq 0 \] where $x \cdot y$ denotes the inner product of $\mathbb{R}^d$, i.e., $X_t(j) \cdot \theta$ is the projection of $X_t(j)$ in the $\theta$ direction. It is conjectured that these derivative martingales figure in the asymptotic law of the point processes \[ \sum_{j=1,\ldots,\mathcal{N}_t} \delta_{(D_t(j), |X_t(j)|-r_t)} \] as $t \to \infty$ where $D_t(j) = X_t(j)/|X_t(j)|$ and $r_t = 2t + \frac{d-4}{2\sqrt{2}} \log t$ is, up to an error of constant order, the median of the maximal displacement of the $d$-dimensional branching Brownian motion. The main result of the paper is that, with probability one, there exists a measurable subset $\Theta$ of the sphere $\mathbb{S}^{d-1}$ of full measure such that $Z_\infty(\theta) = \lim_{t \to \infty} Z_t(\theta)$ exists for all $\theta \in \Theta$. Further, for any bounded measurable function $f$ of $\mathbb{S}^{d-1}$, \[ \lim_{t \to \infty} \int_{\mathbb{S}^{d-1}} Z_t(\theta) f(\theta) \, \sigma(\mathrm{d}\theta) = \int_{\mathbb{S}^{d-1}} Z(\theta) f(\theta) \, \sigma(\mathrm{d}\theta) \quad \text{a.s.} \] with $\sigma$ denoting the uniform distribution on the sphere $\mathbb{S}^{d-1}$. Additionally, $0 < \lim_{t \to \infty} \int_{\mathbb{S}^{d-1}} Z_t(\theta) \, \sigma(\mathrm{d}\theta) < \infty$ a.s.