The maximum of a branching random walk with stretched exponential tails (Q6100140)

Let us given a reproduction law with expectation \(m>1\) and a step size distribution represented by a random variable \(X\). At time \(n=0\) one particle is placed at the origin of the real line. At time \(n=1\) the initial particle splits into \(Z_1\) new particles which move independently of each other and of \(Z_1\). All displacements of particles from their place of birth are independent copies of \(X\). Each particle evolves according to these rules independently of all other particles, and so forth. This procedure generates the Galton-Watson process \(Z=(Z_n)_{n\geq 0}\) with \(Z_0=1\). The paper gives a detailed description of the asymptotic behavior of the position \(M_n\) of the rightmost particle at the time \(n\). The authors assume that the Galton-Watson process \(Z\) is supercritical, \(\mathbb{E}X=0\), \(\mathbb{E}X^2=1\), and there exist \(\lambda>0\), \(r\in (0,1)\) and a function \(a(x)\) such that \(\lim_{x\to\infty}a(x)=a\), \(\mathbb{P}\{X>x\}=a(x)e^{-\lambda x^r}\) for all \(x>0\). The following two conditions appear in the statements of the main results: (1) \(\mathbb{E}e^{s|X|}\) for some \(s>0\) and (2) if \(r>2/3\), \(\lim_{x\to\infty} x^{-\frac{3r-2}{2r-1}}\log \mathbb{P}\{X<-x\}\); if \(r\leq 2/3\), \(\mathbb{E}|X|^k<\infty\), \(\forall k\in\mathbb{N}\). Put \(\alpha=\left(\frac{\log m}{\lambda}\right)^{1/r}\). The paper establishes almost sure convergence of \(\frac{M_n-\alpha n^{1/r}}{n^{2-1/r}}\) for \(r>2/3\) , and convergence in law of \(\frac{M_n-\alpha n^{1/r}}{n^{1/r-1}}\) for \(r\leq 2/3\). It also provides a description of upper and lower space-time envelopes of \(M_n\) in the latter case.