Limit theorems for the maximal tree size of a Galton-Watson forest in the critical case (Q6051965)

Consider a critical Galton-Watson branching process \(\mathcal{F}\) with an initial population consisting of \(N\) individuals, called ancestors, and generic offspring variable \(\xi\). In particular, \(\mathbb{E}[\xi]=1\). Write \[ p_k = \mathbb{P}(\xi=k) = \frac{1}{(k+1)^{\tau}} - \frac{1}{(k+2)^\tau}, \quad k \in \mathbb{N}_0. \] Since \(\mathbb{E}[\xi]=1\) is assumed, this implies \(\sum_{k=1}^\infty kp_k = 1\) and hence determines \(\tau\) to be \(\tau\approx 1.728\). For \(i=1,\ldots,N\), write \(\nu_i(\mathcal{F})\) to denote the total population of the Galton-Watson process stemming from ancestor \(i\), and \(\eta(\mathcal{F}) = \max_{i=1,\ldots,N} \nu_i(\mathcal{F})\) for the maximal size of the \(N\) Galton-Watson trees stemming from the \(N\) different ancestors. The author provides very explicit limit theorems for the distribution function of \(\eta(\mathcal{F})\) conditional given the total population of \(\mathcal{F}\) is \(N+n\), \(N\) ancestors and \(n\) further individuals. Two regimes are considered, namely: \begin{itemize} \item The regime where \(n/N^\tau \to \gamma\) for some constant \(\gamma > 0\) as \(n,N \to \infty\). Here, \(\eta(\mathcal{F})/n\) converges in distribution. \item The regime where \(n/N^\tau \to \infty\) as \(n,N \to \infty\). Then \(\frac{\eta(\mathcal{F})-n}{N^\tau}\) converges in distribution. \end{itemize}