The Seneta-Heyde scaling for the branching random walk (Q2450245)

Consider a branching random walk constructed from a supercritical BGW process \((Z_n)_{n\geq 0}\) and a point process \(\theta\) on the real line in such a way that the children of an individual with position \(x\) have positions distributed according to \(\theta_x\), with \({\mathbf P}(\theta_x\cdot)={\mathbf P}(\theta\{y:y- x\in\cdot\})\). Denote the positions of the \(n\)th generation by \(x_{nj}\), \(j= 1,\dots, Z_n\), and start the process with \(Z_0=1\) and \(x_{00}= 0\). If the process is in the boundary case, i.e., if \[ \sum_j\exp\{-x_{1j}\}=1\quad\text{and}\quad \sum_j x_{1j}\exp\{-x_{1j}\}= 0, \] then the additive martingale \[ W_n:= \sum_j\exp\{-x_{nj}\} \] is known to converge a.s.\ to zero [\textit{J. D. Biggins}, J. Appl. Probab. 14, 25--37 (1977; Zbl 0356.60053)]. Under some additional moment assumptions, the authors prove that, conditioned on non-extinction, \(n^{1/2}W_n\) converges in probability, but not almost surely, to a positive random variable shown to be a.s.\ equal up to a multiplicative constant to the conditional limit \(D_\infty\) of the derivative martingale \[ D_n:= \sum_j x_{nj}\exp\{- x_{nj}\}, \] studied by \textit{J. D. Biggins} and \textit{A. E. Kyprianou} [Adv. Appl. Probab. 36, No. 2, 544--581 (2004; Zbl 1056.60082)].