Spectral radiuses of the Galton-Watson branching processes (Q2369376)

For any irreducible discrete time Markov chain \(\{X_n\}\) with (countable) state space \(E\), there exists a number \(r\in [0,1]\) such that \(\limsup_{n\to\infty} [P(X_n=y\mid X_0=x)]^{1/n} = r\) for each pair \(x,y\in E\). The number \(r\) describes the speed of \(\{X_n\}\) escaping from any state in \(E\) and is called the spectral radius of the process. Let \(\{Z_n\}\) be a one-type Galton-Watson branching process with offspring distribution \((p_0,p_1,p_2,\dots)\) and generating function \(f(s)= \sum_{k=0}^\infty p_ks^k\). Suppose that \(p_0,p_1>0\) and let \(E=\{1\}\) if \(p_0+p_1=1\) and \(E=\{1,2,3,\dots\}\) if \(p_0+p_1<1\). Then \(\{Z_n\}\) is an irreducible Markov chain on \(E\). The authors show this chain has spectral radius \(r=f'(q)\), where \(q\) is the unique solution of the equation \(s=f(s)\) in \([0,1)\). This result is then generalized to the case of a multi-type Galton-Watson branching process. Some examples are given for which the spectral radius is explicitly computable.